A high-throughput protocol for mutationscanning of the BRCA1 and BRCA2 genesHeather L Hondow1, Stephen B Fox1,2, Gillian Mitchell3, Rodney J Scott4, Victoria Beshay1, Stephen Q Wong1,kConFab Investigators and Alexander Dobrovic1,2*
Abstract
Background: Detection of mutations by DNA sequencing can be facilitated by scanning methods to identifyamplicons which may have mutations. Current scanning methods used for the detection of germline sequencevariants are laborious as they require post-PCR manipulation. High resolution melting (HRM) is a cost-effective rapidscreening strategy, which readily detects heterozygous variants by melting curve analysis of PCR products. It is wellsuited to screening genes such as BRCA1 and BRCA2 as germline pathogenic mutations in these genes are alwaysheterozygous.
Methods: Assays for the analysis of all coding regions and intron-exon boundaries of BRCA1 and BRCA2 weredesigned, and optimised. A final set of 94 assays which ran under identical amplification conditions were chosenfor BRCA1 (36) and BRCA2 (58). Significant attention was placed on primer design to enable reproducible detectionof mutations within the amplicon while minimising unnecessary detection of polymorphisms. Deoxyinosineresidues were incorporated into primers that overlay intronic polymorphisms. Multiple 384 well plates were used tofacilitate high throughput.
Results: 169 BRCA1 and 239 BRCA2 known sequence variants were used to test the amplicons. We also performedan extensive blinded validation of the protocol with 384 separate patient DNAs. All heterozygous variants weredetected with the optimised assays.
Conclusions: This is the first HRM approach to screen the entire coding region of the BRCA1 and BRCA2 genesusing one set of reaction conditions in a multi plate 384 well format using specifically designed primers. Theparallel screening of a relatively large number of samples enables better detection of sequence variants. HRM hasthe advantages of decreasing the necessary sequencing by more than 90%. This markedly reduced cost ofsequencing will result in BRCA1 and BRCA2 mutation testing becoming accessible to individuals who currently donot undergo mutation testing because of the significant costs involved.
BackgroundInactivating germline mutations in the BRCA1 andBRCA2 tumour suppressor genes dramatically escalatesthe risk of developing breast and/or ovarian cancer byup to 20 fold [1-4]. Due to the highly penetrant natureof germline mutations within BRCA1 and BRCA2, it isof importance to identify a woman as being a carrier ofa mutation as early intervention measures includingbreast screening and prophylactic bilateral salphingo-
oophorectomy or mastectomy can be offered [5]. Morerecently, it has been recognised that BRCA1 or BRCA2mutant tumours are sensitive to PARP inhibitors andthus rapid and inexpensive BRCA1 and BRCA2 testingmay be of direct clinical utility [6].BRCA1 and BRCA2 are very large genes. BRCA1 has 24
exons (22 of which are protein coding) that code for a1863 amino acid protein while BRCA2 has 27 exons (26coding) that code for a 3418 amino acid protein. Cur-rently, Sanger sequencing is considered as the gold stan-dard for identification of sequence variants withinBRCA1 and BRCA2. However, scanning methods havebeen often employed in order to reduce costs andimprove turn around time. The major disadvantage of
* Correspondence: [email protected] Pathology Research and Development Laboratory, Department ofPathology, Peter MacCallum Cancer Centre, Locked Bag 1, A’Beckett St,Melbourne, Victoria, 8006, AustraliaFull list of author information is available at the end of the article
Hondow et al. BMC Cancer 2011, 11:265http://www.biomedcentral.com/1471-2407/11/265
most scanning methods is that they require post-PCRproduct manipulation which, in addition to the increasedworkload, also carries the potential risks of sample misi-dentification and contamination [7,8].High Resolution Melting (HRM) is a post-PCR method
which enables the detection of sequence variations withinan amplified region of DNA. Using saturating concentra-tions of a fluorescent dye which specifically intercalateswith double stranded DNA, the denaturation behavior ofan amplified region of DNA can be analysed. The dyedissociates from the double stranded DNA as it dena-tures into single stranded DNA and thus the melting canbe monitored using the decrease in fluorescence. Hetero-zygous sequence variations are readily detected due tothe formation of heteroduplexes between variant andwildtype strands that then have a characteristic earlymelting profile [9,10].HRM has the major advantage over other pre-sequen-
cing scanning methods in that it is performed in a “closedtube” system. This eliminates the risk of post-PCR pro-duct contamination during scanning while also reducingprocessing time (especially when the PCR and HRM areperformed within the one instrument as in this study),resulting in improved turn around times. HRM has effec-tively replaced the previously most commonly used scan-ning method, denaturing high pressure liquidchromatography (DHPLC). It has better sensitivity andspecificity for the detection of variants than DHPLC[7,8,11].In this communication, we report the development of
an HRM-based assay system for mutation detectionwithin BRCA1 and BRCA2 using a 384 well plate formatboth to facilitate the detection of mutations and to enablehigh throughput scanning. As it is difficult to distinguishheterozygosity for SNPs from heterozygosity of othersequence variants, we have employed strategies for SNPminimisation within amplicons. We also introduced theuse of deoxyinosine residues in HRM primers to enablethe siting of primers over clinically insignificant intronicSNPs. The assay system is a robust BRCA1 and BRCA2mutation scanning protocol that has had the most exten-sive validation so far reported.
ResultsAmplicon design principlesAmplicons were selected in order to analyse the entirecoding sequence and the intron-exon boundaries ofBRCA1 and BRCA2. Typically, one amplicon wasdesigned per exon. However, for longer exons or exonswith more complex melting domains, two or more over-lapping amplicons were chosen.Extensive in silico analysis was performed in order to
identify amplicons that would be suitable for both PCRand HRM. The DNA melting prediction software
‘Poland’ [12] was used to choose amplicons which pre-ferably had a single melting domain. In some cases,amplicons with multiple domains were selected to keepthe overall number of amplicons low. This was espe-cially the case where the amplicon was short. Wheredouble melting domains were unavoidable e.g. BRCA2exon 11Q, the tested mutations were readily detectable(Additional file 1 Figure S1).In germline DNA, all pathogenic BRCA1 and BRCA2
mutations will exist in heterozygous form giving rise toheteroduplexes that enhance variant detection by HRM.Previous HRM studies have reported 100% sensitivity fordetection of heterozygous sequence variations in PCRproducts up to 435 bp [13-15]. All of the amplicons inthis study were less than 405 bp. Although the homozy-gous genotypes for some of the polymorphisms werereadily distinguishable e.g. the c.2612C>T in BRCA1exon 11H (Additional file 2 Figure S2), others were not.This was not considered a problem as our aim was toidentify heterozygous changes rather than to genotypeexisting high frequency polymorphisms.
Primer design principlesIn order for all of the assays to perform under the samePCR and HRM conditions, primers were designed to havean annealing temperature of 64-67°C using OligoCalc [16].Primers were designed to minimise primer-dimer interfer-ence and were screened to ensure specificity to the targetsite using Amplify v3.1 [17]. Attention was paid to the 3’ends of the primers ensuring that they were neither tooGC rich or AT rich.Intronic polymorphisms within an amplicon decrease
the specificity of HRM assays for mutation detection asthey produce heterozygous melting profiles and thusrequire sequencing to distinguish them from true muta-tions. Primers were placed as close to exon boundaries aspossible while still leaving at least 5 intronic bases toidentify the most likely splicing mutations. This alsoassisted in keeping the amplicon size lower. However,where there were known or suspected pathogenic intro-nic variants close to the exons listed in the Breast CancerInformation Core Database (i.e. BRCA1 c.213-11T>G,BRCA1 c.5194-12G>A and BRCA2 c.426-12_8del5), theprimers were moved further into the intron to enable thedetection of these variants [18].A critical consideration in primer design for mutation
screening is that primers should not be placed over knownpolymorphisms, even if they are comparatively rare. Thismay lead to non-amplification of the polymorphic alleleand coexisting pathogenic mutation if in a cis relationshipwith the polymorphism or false homozygosity of the inac-tivating mutation if it exists in a trans relationship withthe polymorphic allele [19,20]. Care was thus taken toidentify known single nucleotide polymorphisms (SNPs)
Hondow et al. BMC Cancer 2011, 11:265http://www.biomedcentral.com/1471-2407/11/265
Page 2 of 11
using the Basic Local Alignment Search Tool (BLAST)[21]. Care was also taken to ensure that the primer setsdid not non-specifically amplify the BRCA1 partial pseu-dogene where the homology extends into BRCA1 exon 2[22].When common or rare polymorphisms were present
in an intronic region that was otherwise optimal for pri-mer placement, we incorporated a deoxyinosine residue(dI) at the position of the polymorphism. Incorporatinga dI into the primer at the site of the polymorphismallows equal amplification of both the wildtype and thevariant allele [23]. During amplification, complementarybases will be inserted randomly producing a productthat is indistinguishable from wildtype by HRM as dIwill not form heteroduplexes regardless of which base itis paired with. Consistent with this, we found that incor-porating a dI into the primer at the location of the poly-morphism was a good strategy for eliminating thedetection of the polymorphism while giving clean HRMprofiles suitable for both mutation detection and down-stream sequencing of mutations. Polymorphisms thatwere replaced by dI were, BRCA2 c.-26G>A (rs1799943)which flanks the coding region of exon 2, BRCA2c.7806-14T>C (rs9534262) and BRCA1 c.80+13A>G.Amplification of both alleles of the polymorphism allowsthe detection of all coexisting mutations while maskingthe SNP. Figure 1 shows the example of detection ofboth the BRCA2 c.26delC mutant allele and the wildtype allele where a dI in the primer has been used tomask the c.-26 G>A polymorphism.
Protocol designIn HRM analysis, multiple samples for each ampliconshould be screened at the same time in order to enablethe ready detection of mutant sequences relative to mul-tiple wild-type sequences as more samples act to
dampen stochastic variation. As most samples are wild-type for any given amplicon, multiple samples will makeit easy to distinguish a mutation from the wildtype var-iation. The use of 384 well plates facilitates the use ofmultiple samples. We chose a batch size of either 10 (6plates) or 22 samples (12 plates). A known wild typecontrol that was homozygous for all the common SNPsand a no-template control were also included. Theamplicons were clustered into discrete areas on theplate to minimise the temperature variation for eachamplicon (Additional file 3 Table S1 shows the layout).Each primer set was initially tested using the standard
mastermix and PCR and HRM program, after which pri-mer sets requiring additional optimisation were testedusing the simplified primer matrix. Only primer setswhich produced specific products as assessed on anagarose gel and which both amplified efficiently andmelted in an acceptable profile were used for mutationtesting. Otherwise, new primer sets were designed forthe region. In addition, primers needed to give goodmelt curves without any non-specific products.The analysis was performed on a combined PCR and
HRM instrument allowing us to analyze HRM data in thecontext of the PCR amplification information. Duringanalysis of the amplification for each amplicon, any out-lying late amplifying replicates that were often associatedwith false positive alterations in melting profile wereremoved from the analysis.
Optimizing the set of HRM amplicons against a panel ofknown sequence variantsIn the final protocol, 36 amplicons were used to analyzeBRCA1 while 58 amplicons were used to analyze BRCA2(Additional file 4 Table S2). Initially, a provisional set ofassays were designed. For these, a large panel of muta-tion positive and other sequence variant controls were
Figure 1 Detection of both BRCA2 c.26delC mutant and wild type alleles where deoxyinosine is used at the c.-26G>A SNP. The red andblue samples both contain the c.26delC deletion but differ according to their genotype for the BRCA2 c.-26G>A polymorphism. The red profile ishomozygous wildtype for the polymorphism. The blue profile is heterozygous for the polymorphism. The left panel shows the difference curveswhere the baseline comprises both wildtype (grey) and heterozygous (green) genotypes. The right panel shows the corresponding melting peakcurves. The difference curves are independent of the SNP meaning that both alleles are equally amplified using the deoxyinosine containingprimers
Hondow et al. BMC Cancer 2011, 11:265http://www.biomedcentral.com/1471-2407/11/265
Page 3 of 11
used to test the ability of the amplicons to identifysequence variants. The controls were analysed alongsidemultiple wild type controls derived from women thathad previously tested negative for sequence variants.Some amplicons required further optimisation of the
primer concentration which was determined using arange of concentrations of the forward and reverse pri-mers in a primer matrix [24]. Agarose gels were run foreach primer set to ensure that the correct size productwas produced and that there were no non-specific pro-ducts or primer dimers formed.Mutation detection was carried out using all the three
visualisations of the raw data (normalised and tempera-ture shifted melting curves, normalised and temperatureshifted difference curves, and the negative first derivativecurves which give a melting peak curve). Melting peakcurves are independent of normalisation and tempera-ture shifting. However in general, the difference curveview was the most useful. Normalising the data at thepre-and post-melt phases of HRM allows samples to bedirectly compared.Surprisingly since all heterozygous mutations might be
expected to have pronounced heteroduplexes, somemutations, especially single base insertions and dele-tions, gave rise to more subtle shifts. Previous commu-nications have also noted the difficulty in detectingsingle base insertion or deletion mutations [25,26]. Wehave found this to be most marked when the single baseinsertion or deletion exists within extended runs of asingle nucleotide. Such variations create little Tm differ-ence and are consequently more subtle with HRM.While we did experiment with additives such as DMSOwith some success for difficult amplicons, we finallyfocused on amplicon design to maintain consistency inmastermix set-up.
Figure 2 shows an example where the c.2885delAmutation in BRCA1 is less apparent than the othersequence variants, particularly when the melting peakcurves are examined. Nevertheless the mutation is stillobviously different from the wildtype. While singlenucleotide changes and multiple base insertions anddeletions are usually best detected using the normalisedand temperature shifted melting curves, single baseinsertions and deletions are better detected on a differ-ence curve relative to a wildtype control.In some amplicons, certain mutations were initially not
detected by the HRM assays. The exons that proved mostproblematic were BRCA1 exon 7 and BRCA2 exons 3, 11and 15. For example, in BRCA1 exon 7, both the singlebase insertion c.329insA and single base deletion muta-tion c.302-2delA were not detected with the originalamplicon. As a result, new amplicons were designed,optimised and validated in order to also detect the muta-tions and checked against the other mutation controls.For BRCA2 amplicon 11I, we were not initially able todetect the c.4512insT which inserted an extra T into arun of 6 Ts and thus included an extra amplicon (11Is)that covered a shorter region and was able to resolve thismutation (data not shown).Another solution was to divide problematic amplicons
into 2 overlapping shorter amplicons making the meltingprofile within each amplicon less complex based onPoland analysis (Figure 3). By decreasing the size of theamplicon, we could detect all available mutations, includ-ing those that are located within repeat regions (Figure 4).A large number of positive controls were tested for
BRCA1 (184) and BRCA2 (256) (Additional file 5 TableS3). While most amplicons had multiple mutation con-trols (mean = 4.25), a few amplicons did not possessany mutation controls. These amplicons were BRCA1
Figure 2 Detection of different mutations within the same amplicon (BRCA1 11I region). The left panel shows the difference curves andthe right panel shows the corresponding melting peak curves. All mutations produce obvious biphasic melting which is caused by the earliermelting of the heteroduplexes. The normalised and temperature-shifted difference plots in the left panel allow easy detection of thec.2863delTCATC (navy), c.2800C>T (red) and c.2885delA (green) relative to the wildtype HRM profile (grey). The melting peak curves in the rightpanel show that the c.2885delA is the most subtle mutation in that there is a minimal early melting heteroduplex component compared to thec.2863delTCATC and c.2800C>T mutations.
Hondow et al. BMC Cancer 2011, 11:265http://www.biomedcentral.com/1471-2407/11/265
Page 4 of 11
exons 9 and 22, BRCA2 exons 11J2, exon 12, exon 26and exon 27A. However, during subsequent blindedtesting, mutations in all of these exons except 27A weredetected. These and other new variants that weredetected are detailed in Additional file 6 Table S4.
Validation of the assaysIn addition to the testing of the individual amplicons,two sets of validations using full HRM screens were
performed. In these validations, we undertook a deliber-ate strategy of sequencing all reactions with minordeviations from wildtype i.e. those that were likely to be“false positives”. As discussed, some pathogenic muta-tions can produce subtle changes to the melting beha-viour. All reactions which were suspicious includingthose that had one aberrant replicate were sequenced.Thus, likely false positives were targeted for sequencingas the consequences of inadvertently missing a mutation
Figure 3 The influence of amplicon choice on mutation detection. The top panels show difference curves and melting peak curves for theoriginal BRCA1 exon 7 amplicon which did not readily detect single base insertions or deletions. While the c.427G>T (red) and the c.314A>G(green) mutations are readily detectable in both visualisations, the pathogenic c.329insA (navy) and c.302-2delA (yellow) single base pair insertionand deletion mutations were difficult to detect as they melt like the wildtype controls (grey). The original amplicon was then divided into twooverlapping amplicons. The bottom panels show difference curves and melting peak curves for BRCA1 amplicon exon 7A which detects thesingle base insertion and deletion. The c.329insA (navy) and c.302-2delA (yellow) clearly differ from the wildtype controls (grey) in the shorteramplicon. The BRCA1 c.314A>G (dark green) mutation is also readily detectable in both visualisations.
Figure 4 Detection of an insertion within a long nucleotide repeat by HRM (BRCA2 exon 23). The mutation c.9097insA (green) is aninsertion of an adenine nucleotide into an 8 adenine repeat. Like the other mutations here; c.9117G>A (red) and c.9117+1G>A (navy); it isreadily distinguishable from the wildtype (grey) in this 264 bp amplicon.
Hondow et al. BMC Cancer 2011, 11:265http://www.biomedcentral.com/1471-2407/11/265
Page 5 of 11
are more serious than calling a false positive. We alsosequenced all those variants that were likely to be SNPs.In the first set, 266 samples which were sent for diag-
nostic testing by bidirectional Sanger sequencing wereconcurrently screened by the first version of our fullHRM screen. The operator was blinded to the results ofthe mutation sequencing until the samples had beenscored as to whether they contained variants by HRM.Subsequently the results of HRM and sequencing werecompared. All variants including heterozygous SNPswere considered in the analysis.22,797 amplicons were analysed and 7.5% were identi-
fied as requiring sequencing to identify the variant while92.5% were considered to be wildtype. All 35 pathogenicmutations (13% of DNA samples) identified by sequen-cing were also detected by the HRM screening. The cal-culated sensitivity for the detection of heterozygousvariants (including SNPs) was 99.8% (1595/1599) whilethe specificity was 99.4% (21,070/21,192). The positivepredictive value was 92.9% while the negative predictivevalue was 99.98%. At this stage, the BRCA1 c.5074+3A>C, BRCA2 c.1834G>A, c.2971A>G, c.3807T>C andc.9458G>C variants were not detectable but subsequentamplicon redesign as used in the final protocol led totheir detection and an effective sensitivity of 100%.A second validation was performed by retrospectively
re-screening 118 archival samples which were previouslyreported as negative for germline inactivating mutationswithin BRCA1 and BRCA2 with the final version of theprotocol. These samples had previously undergone test-ing using a combination of testing procedures includingprotein truncation testing (PTT) and partial Sangersequencing. Some of these samples had only been testedfor one of the genes or had just undergone PTT testingfor mutations in large exons. In the second validation,11,092 fragments were analysed and 92.7% of amplicons
were called wildtype. 7.3% of amplicons contained var-iants that required further investigation by sequencing.Sensitivity was 100% while specificity was 99.75%. Thepositive predictive value was 98.98% while the negativepredictive value was 100%. The positive predictive valuewas decreased due to the deliberate strategy of alsochoosing reactions with minor deviations from wildtypethat were likely to be “false positives” by HRM.Once again, the positive predictive value was
decreased due to the detection of “false positives” byHRM that were identified as wildtype by sequencing.Eighteen amplicons were “falsely positive” by HRM.Importantly, no amplicons were falsely negative byHRM.
Mutations co-existing with polymorphismsIn some cases, mutations may be present close to acommon or rare polymorphism. De Juan et al. (2009)raised the theoretical concerns that pathogenic muta-tions which coexist with polymorphisms may distort theHRM curve and make it appear like a normal sequence[27]. We consider this extremely unlikely as greaterinstability (early melting) would be caused by multiplepopulations of heteroduplexes formed as a result of thetwo sequence variants. In support of this, we found sev-eral examples of a mutation coexisting with a commonpolymorphism and resulting in a greater differencebetween the melting curves of the doubly variant poly-morphic sample and the wildtype (Figure 5).
Detection of mutations adjacent to the 3’ end of theprimersIt has been argued by some authors that mutations closeto the primers are more difficult to detect than morecentrally placed mutations. While it has been reportedthat the ability to detect mutations decreases as the
Figure 5 Two variations within the same sample result in greater HRM differences. (BRCA1 exon 13). The left panel shows the differenceplot where 3 samples have been compared to the wildtype controls (grey). The green profile represents the a heterozygote for commonpolymorphism c.4308T>C while the blue represents the mutation c.4327C>T. The presence of both sequence variants (red) results in greaterinstability than each individual sequence variation due to the double mismatch in heteroduplexes. The right panel with melting peak curvesshow the complex melting nature of two co-existing sequence variants (red).
Hondow et al. BMC Cancer 2011, 11:265http://www.biomedcentral.com/1471-2407/11/265
Page 6 of 11
distance between the primer and the mutation decreases[8,28], we were able to detect mutations immediatelyadjacent to the 3’ end of the primer; an example beingthe BRCA2 c.1365A>G polymorphism in amplicon exon10C (Figure 6). Other publications have confirmed thatthe location of the sequence variation within the ampli-con does not affect the ability of HRM to detect the var-iation [9,29]. Van der Stoep et al. (2009) also reporteddetecting mutations up to 2 bp from the end of the pri-mer [25].
DiscussionCurrent dideoxy (Sanger) sequencing based methods formutation identification are expensive and labor intensiveeven with a high degree of automation. Scanning meth-odologies can markedly reduce the amount of sequen-cing by identifying the limited number of PCRamplicons carrying variant sequences. High resolutionmelting has become the method of choice for scanningas it can be performed immediately after PCR amplifica-tion in a seamless closed tube protocol on a singleinstrument. It is also preferable to other scanning meth-ods as it does not involve out-of-tube manipulation ofthe PCR product thereby reducing the potential for PCRcontamination. Well designed and validated HRM assaysdetect practically every sequence variant [30].Our aim was to develop a high throughput HRM pro-
tocol in order to enable BRCA1 and BRCA2 testing tobe performed at considerably lower cost without com-promising sensitivity or specificity. The cost reduction issignificant as the number of amplicons to be sequencedis reduced by more than 90%. This involved sequencingof all variants. The majority of these turned out to beSNPs as expected. Within a research setting, costs couldbe further reduced by not sequencing those samplesfrom amplicons spanning a common SNP where themelting curves were identical to those of the SNP.An extensive validation process confirmed that HRM is
a highly effective screen for the detection of heterozygous
sequence variants within the BRCA1 and BRCA2 genes.Since pathogenic germline mutations always exist in aheterozygous state within germline DNA and homozy-gous germline variants only exist for non-clinicallyimportant variants such as polymorphisms, HRM is thusideal for minimizing the burden of sequencing.In the past, a combined pre-sequencing screening pro-
tocol of PTT and DHPLC followed by sequencing ofamplicons found to possess sequence variants was oftenused. DHPLC was formerly the gold standard for variantscanning [31]. Comparative studies have shown thatHRM has superior sensitivity and specificity thanDHPLC [7,8]. Whereas HRM is able to detect all hetero-zygous mutations and some homozygous mutationswithin amplicons with complex melting domains,DHPLC sensitivity and specificity can be challengedwithin multiple melting domains. DHPLC also requiresthat the column temperature must also be empiricallyoptimised for each amplicon whereas; we have devel-oped HRM so that all amplicons undergo the same PCRand HRM reaction conditions. Thus, multiple HRManalyses can thus be carried out in parallel contrastingwith DHPLC in which analyses are sequential.HRM has been used as a rapid screen for BRCA1 and
BRCA2 founder mutations. Dufresne et al (2006) firstreported HRM for the detection of the 3 common Ash-kenazi founder mutations using short amplicons [32].We subsequently established that long amplicons couldbe successfully used for the Ashkenazi mutations [33].This enabled the detection of other mutations, and thisprinciple underlies the experimental design in this paperwhere it is often necessary to screen comparatively longstretches of DNA in order to efficiently scan the fullcoding sequence of the genes. Other populations studiedfor founder mutations by HRM include Spaniards,Greeks and Southern Chinese [27,34,35].Full screening by HRM of exons including intron-exon
boundaries has been reported for BRCA1 [25], and forBRCA1 and BRCA2 [26,36]. We have independently
Figure 6 Detection of a sequence variation immediately adjacent to the 3’ end of a primer. The class 1 SNP (BRCA2 c.1365A>G) (red) isreadily detectable at the position immediately adjacent (1 base pair) to the 3’ end of the reverse primer in this 232 bp amplicon (BRCA2 exon 10C).
Hondow et al. BMC Cancer 2011, 11:265http://www.biomedcentral.com/1471-2407/11/265
Page 7 of 11
developed a comprehensive HRM protocol for fullscreening of exons and intron-exon boundaries for bothgenes. We aimed to design all the PCR assays so theycould be amplified under identical conditions using aminimal amount of DNA in a high throughput (384well) approach. Previous communications used 96 wellplates and the layout and ergonomics were notdescribed [26,36].Each assay was tested using as many positive mutation
controls as available to us. Where sequence variations werenot immediately detectable as occurred in a few cases, newprimers were optimised for the detection of that variationin addition to the other positive controls present within theamplicons. Extensive testing of the amplicons with knownmutations and redesign where there were detection diffi-culties, resulted in a set of primers that enabled the detec-tion of 100% of mutations in the panel.Two sets of blinded validation experiments were per-
formed, making this the most extensively validated HRMprotocol described. Firstly, validation was performed by aprospective blinded screening of 266 samples sent in fordiagnostic mutation testing using Sanger sequencing. Anadditional retrospective blinded validation was performedby screening 118 previously tested samples in which nomutations had been detected at the time of either full orpartial testing. All previously described sequence variantscould be detected and a pathogenic mutation in a pre-viously untested region was detected.HRM screening is most cost-effective in the absence
of a large number of SNPs. The relatively low amountof SNPs in BRCA1 and BRCA2 make HRM a suitablemutation scanning technique for these genes. We usedappropriate primer design (including the use of inosines)and placement to further minimise the SNPs withinamplicons. Nevertheless those SNPs that occur withincoding sequences could not be eliminated. Whilesequencing burden is significantly reduced via pre-sequencing HRM screening, approximately 7% of ampli-cons still require sequencing which is largely due to thedetection of clinically non-significant polymorphisms. Inour confirmatory sequencing, 97% of the identifiedsequences were SNPs, entailing an average of 7 ampli-cons that needed to be screened per patient.Others have incorporated SNP genotyping using unla-
belled probes as part of their protocol [37,38]. Althoughwe have shown earlier that samples that are heterozy-gous for both a polymorphism and a mutation can bedistinguished from those carrying a polymorphismalone, we believe that it is safer to sequence all sequencevariants in a diagnostic scenario.
ConclusionsOur protocol comprises 94 PCR amplicons that coverthe complete coding region along with intron-exon
boundaries of BRCA1 and BRCA2. The amplicons aredesigned to be run under the same PCR and HRM con-ditions. A 384 well format allows the greatest number ofHRM assays to be performed at the one time as well asallowing more reliable detection of mutations. Theclosed-tube format used here has clear advantages overother screening technologies which require post-PCRproduct manipulation as there is a complete eliminationof all risk of PCR product contamination along withconsiderable reduction in manual handling [8,39,40]. Itwas shown that the incorporation of deoxyinosine intothe primers was useful in both allowing flexible primerplacement and reducing the amount of subsequentsequencing necessary.The protocol has applications within all diagnostic and
research projects that have budget limitations. It can berun at a fraction of the cost of a full sequencingapproach as the sequencing burden is relieved by morethan 90%. This marked reduction in cost will enableBRCA1 and BRCA2 mutation detection to be morewidely used especially for those who have a positivefamily history for breast or ovarian cancer but who donot meet current algorithms for mutation testing. Theassay design principles in this paper are also relevant tothe HRM screening of other germline mutations.
MethodsPatient DNA SamplesSamples and control DNA for the HRM optimisationand targeted mutation detection were obtained from theDiagnostic Molecular Pathology laboratory at the PeterMacCallum Cancer Centre and from the Kathleen Cun-ningham Foundation Consortium for Research intoFamilial Breast Cancer (kConFab). Samples for theblinded prospective validation (266) and for the blindedretrospective validation (118) were obtained from theDiagnostic Molecular Pathology laboratory at the PeterMacCallum Cancer Centre. Additional blinded BRCA2mutation controls (18) and blinded DNA samples (10)were obtained from the Hunter Area Pathology Service,John Hunter Hospital and Royal Melbourne Hospital,respectively. This study was conducted under guidelinesapproved by the Peter MacCallum Ethics of HumanResearch Committee (approval number 03/90). All indi-viduals had previously consented for germline BRCA1and BRCA2 testing.
Setting up of reactionsEpMotion 5075 (Eppendorf AG, Eppendorf, Germany)programs were designed for the dispensing of the mas-termix and samples into pre-designated wells of a 384well thermoplate (Roche Diagnostics, Penzberg, Ger-many). All reactions were set up in duplicate. Two alter-native programs were used; a 6 plate program with 16
Hondow et al. BMC Cancer 2011, 11:265http://www.biomedcentral.com/1471-2407/11/265
Page 8 of 11
assays per plate screening 10 samples plus 1 wildtypecontrol and 1 no-template control and a 12 plate pro-gram with 8 assays per plate screening 22 samples plus1 wildtype control and a no-template control. The platelayouts are set out in Additional file 3 Table S1.Following preparation of the mastermix, 7 uL was ali-
quotted into each pre-designated well. As multi-dispen-sing on the EpMotion is restricted to volumes equal orgreater than 3 uL, the DNA (10 ng per reaction) wasadjusted with PCR grade H2O to 3.3 ng/μl. The totaltime taken to set up each plate was less than 1 hour asmastermix dispensation and DNA addition eachrequired 20 minutes.
PCR and HRM conditionsSamples were run in duplicate using 10 ng of genomicDNA, 250 nM of forward and reverse primers, master-mix and PCR grade water in a total reaction volume of10 uL. Either LightScanner Master Mix (Idaho Technol-ogy, Salt Lake City, UT) or TrendBio Master Mix (Tren-dBio Pty Ltd, Melbourne, Australia), supplemented byLCGreenPlus™ Hi-Res Melting Dye (Idaho Technology)were used. The primers for BRCA1 exons 11I, 11M, 17and 18, and BRCA2 exon 16 were used at 500 nM finalconcentration. The primer sequences used are listed inAdditional file 4 Table S2.PCR and HRM was performed on the LightCycler 480
(Roche Diagnostics). Template amplification conditionsincluded an activation step of 10 minutes at 95°C fol-lowed by 45 denaturation cycles of 95°C for 10 seconds,annealing for 10 seconds comprising 10 cycles of atouchdown from 65 to 55°C at 1°C/cycle followed by 35cycles at 55°C, and extension at 72°C for 30 seconds.Prior to the HRM, a heteroduplex forming step involvedheating the PCR products to 95°C for 1 minute and arapid cooling to 45°C for 1 minute. HRM was per-formed from 72°C through to 95°C at a temperaturegradient of 1°C per second, acquiring 30 data pointsper °C.
HRM analysisThe melting curves were normalised at the pre-melt(100% fluorescence) and post-melt (0% fluorescence)stages using the supplied software temperature shifting[13] was used to compensate for temperature variationbetween wells and enabled samples with similar dena-turation behaviour to be grouped. While there may besome loss of information with increasing temperatureshifts, in particular, loss of detection of homozygous var-iations, visualisation of heterozygous profiles, particu-larly those of single base insertions and deletions isenhanced at greater temperature shifts. The defaultthreshold used in our study was 5. However, the
threshold can be lowered when an ambiguous sample isbeing examined. We used a default sensitivity setting of0.70.Samples with aberrant melting behavior were chosen
as discussed in the Results section. A conservativeapproach was taken which maximises sensitivity at theloss of specificity i.e. some false positive calls weresequenced to avoid missing some of the more subtlemutations.
SequencingChosen samples were directly sequenced from a 1/35dilution of the HRM product using the BigDye Termi-nator v3.1 cycle sequencing kit (Applied Biosystems,Foster City, CA). Following manual ethanol precipitationand clean up, Hi-Di™ Formamide (Applied Biosystems)denaturation, the samples were analysed on the ABI3730 DNA sequencer (Applied Biosystems). The result-ing sequence data was analyzed with Sequencher soft-ware, version 4.9 (Gene Codes, Ann Arbor, MI).
Additional material
Additional file 1: Figure S1: Detection of mutations within a doublemelting domain. Despite the very clear double melting domain, bothmutations; BRCA2 c.6743del13bp (blue) and BRCA2 c.6821G>T (green) arereadily differentiated from the wildtype (grey) in this 221bp amplicon.
Additional file 2: Figure S2: Visualisation of all three genotypes ofthe BRCA1 c.2612C>T SNP. The wildtype C/C homozygote is distinctfrom the T/T homozygote. The heterozygote has a broader melting peakwhich is due to the combined melting peaks of the homoduplex andheteroduplex populations.
Additional file 3: Table S1: Reaction layout. This table shows thelayout used for each of the 12 plates in a 22 sample run. Each plate canrun 8 assays with 22 test samples, a wildtype control and a no templatecontrol.
Additional file 4: Table S2: PCR primer sequences. This table showsthe final set of primers used. M13 sequences, where used, are indicatedin bold.
Additional file 5: Table S3: List of known variants used in testing.These represent the known mutations that were tested for BRCA1 andBRCA2.
Additional file 6: Table S4: List of previously untested variantsfound during validation. These are the previously undetected variantsthat were detected during blinded testing.
AcknowledgementsThis project was supported by a Priority Driven Collaborative CancerResearch Scheme grant from Cancer Australia partnered by the NationalBreast Cancer Foundation and the Prostate Cancer Foundation of Australiato AD, SF and GM and a New Technology grant from the VictorianDepartment of Human Services to SF, GM and AD. We thank TrentWarburton from Trendbio Sciences for technical discussion. We especiallythank our consumer representative, Gerda Evans for her support. We thankHongdo Do for his critical reading of the final version of the manuscript. Wethank Cliff Meldrum (Peter MacCallum Cancer Centre) and Margie Smith andEmanouil Sigalas (Royal Melbourne Hospital) for DNA samples used in thisstudy. We also thank Amber Willems and Heather Thorne from kConFab forpositive control DNA samples. We thank the kConFab research nurses and
Hondow et al. BMC Cancer 2011, 11:265http://www.biomedcentral.com/1471-2407/11/265
staff, the heads and staff of the Family Cancer Clinics, and the Clinical FollowUp Study for their contributions to the kConFab resource, and the manyfamilies who contribute to kConFab. kConFab is supported by grants fromthe National Breast Cancer Foundation, the National Health and MedicalResearch Council of Australia, the Queensland Cancer Fund, the CancerCouncils of New South Wales, Victoria, Tasmania and South Australia, andthe Cancer Foundation of Western Australia.
Author details1Molecular Pathology Research and Development Laboratory, Department ofPathology, Peter MacCallum Cancer Centre, Locked Bag 1, A’Beckett St,Melbourne, Victoria, 8006, Australia. 2Department of Pathology, TheUniversity of Melbourne, Parkville, Victoria, 3010, Australia. 3Familial CancerCentre, Peter MacCallum Cancer Centre, Locked Bag 1, A’Beckett St,Melbourne, Victoria, 8006, Australia. 4School of Biomedical Sciences,University of Newcastle, New South Wales, 2308, Australia.
Authors’ contributionsHH carried out and analysed the molecular genetic studies and drafted theoriginal manuscript. AD conceived the study, and participated in its designand coordination, took the manuscript to completion, and wrote therevision. AD and HH designed all the amplicons used in the final protocol.VB participated in the comparison of the blinded HRM assays with thesequencing results. SQW assisted in the data interpretation, analysis and inthe writing of the revisions. SBF, GM and RJS participated in the design ofthe study and supply of specimens. kConFab supplied specimens withknown mutations as positive controls. All authors read and approved thefinal manuscript.
Received: 25 January 2011 Accepted: 24 June 2011Published: 24 June 2011
Liu Q, Cochran C, Bennett LM, Ding W, Bell R, Rosenthal J, Hussey C, Tran T,McClure M, Frye C, Hattier T, Phelps R, Haugen-Strano A, Katcher H,Yakumo K, Gholami Z, Shaffer D, Stone S, Bayer S, Wray C, Bogden R,Dayananth P, Ward J, Tonin P, Narod S, Bristow PK, Norris FH, Helvering L,Morrison P, Rosteck P, Lai M, Barrett JC, Lewis C, Neuhausen S, Cannon-Albright L, Goldgar D, Wiseman R, Kamb A, Skolnick MH: A StrongCandidate for the Breast and Ovarian Cancer Susceptibility Gene BRCA1.Science 1994, 266:66-71.
2. Wooster R, Neuhausen SL, Mangion J, Quirk Y, Ford D, Collins N, Nguyen K,Seal S, Tran T, Averill D, Fields P, Narod S, Lenoir GM, Lynch H, Feunteun J,Devilee P, Cornelisse CJ, Menko FH, Daly PA, Ormiston W, McManus R,Pye C, Lewis CM, Cannon-Albright LA, Peto J, Ponder BAJ, Skolnick MH,Easton DF, Goldgar DE, Stratton MR: Localization of a breast cancersusceptibility gene, BRCA2, to chromosome 13q12-13. Science 1994,265:2088-2090.
3. Narod SA, Foulkes WD: BRCA1 and BRCA2: 1994 and beyond. Nature RevCancer 2004, 4:665-676.
4. Antoniou A, Pharoah PD, Narod S, Risch HA, Eyfjord JE, Hopper JL,Loman N, Olsson H, Johannsson O, Borg A, Pasini B, Radice P, Manoukian S,Eccles DM, Tang N, Olah E, Anton-Culver H, Warner E, Lubinski J,Gronwald J, Gorski B, Tulinius H, Thorlacius S, Eerola H, Nevanlinna H,Syrjäkoski K, Kallioniemi OP, Thompson D, Evans C, Peto J, Lalloo F,Evans DG, Easton DF: Average risks of breast and ovarian cancerassociated with BRCA1 or BRCA2 mutations detected in case seriesunselected for family history: a combined analysis of 22 studies. Am JHum Genet 2003, 72:1117-1130.
5. Kauff ND, Satagopan JM, Robson ME, Scheuer L, Hensley M, Hudis CA,Ellis NA, Boyd J, Borgen PI, Barakat RR, Norton L, Castiel M, Nafa K, Offit K:Risk-reducing salpingo-oophorectomy in women with a BRCA1 orBRCA2 mutation. N Engl J Med 2002, 346:1609-1615.
6. Rubinstein WS: Hereditary breast cancer: pathobiology, clinicaltranslation, and potential for targeted cancer therapeutics. Fam Cancer2008, 7:83-89.
7. Chou LS, Lyon E, Wittwer CT: A comparison of high-resolution meltinganalysis with denaturing high-performance liquid chromatography formutation scanning: cystic fibrosis transmembrane conductanceregulator gene as a model. Am J Clin Pathol 2005, 124:330-338.
8. Sadr-Nabavi A, Hoffman M, Weilke C, Sun YL, Nevinny-Stickel-Hinzpeter C:Mutation Scanning Using High Resolution Melting or dHPLC: aPerformance Comparison Study. Biochemica 2009, 4:30-33.
9. Reed GH, Wittwer CT: Sensitivity and specificity of single-nucleotidepolymorphism scanning by high-resolution melting analysis. Clin Chem2004, 50:1748-1754.
10. Montgomery J, Wittwer CT, Kent JO, Zhou L: Scanning the cystic fibrosistransmembrane conductance regulator gene using high-resolution DNAmelting analysis. Clin Chem 2007, 53:1891-1898.
11. Ravnik-Glavac M, Atkinson A, Glavac D, Dean M: DHPLC screening of cysticfibrosis gene mutations. Hum Mutat 2002, 19:374-83.
12. The Poland thermal denaturation profile program. [http://www.biophys.uni-duesseldorf.de/local/POLAND//poland.html].
17. Amplify 3 program for MacOS X. [http://engels.genetics.wisc.edu/amplify/].18. NHGRI Breast Cancer Information Core Database. [http://research.nhgri.
human BRCA1 knockout caused by mispriming during polymerase chainreaction: implications for genetic testing. Genes Chromosomes Cancer2001, 31:96-98.
20. Solano AR, Dourisboure RJ, Weitzel J, Podesta EJ: A cautionary note: falsehomozygosity for BRCA2 6174delT mutation resulting from a singlenucleotide polymorphism masking the wt allele. Eur J Hum Genet 2002,10:395-397.
21. Single Nucleotide Polymorphism BLAST. [http://www.ncbi.nlm.nih.gov/SNP/snpblastByChr.html].
22. Brown MA, Xu CF, Nicolai H, Griffiths B, Chambers JA, Black D, Solomon E:The 5’ end of the BRCA1 gene lies within a duplicated region of humanchromosome 17q21. Oncogene 1996, 12:2507-2513.
23. Shen Z, Liu J, Wells RL, Elkind MM: Direct sequencing with highlydegenerate and inosine-containing primers. Methods Mol Biol 1996,65:111-118.
24. Mikeska T, Dobrovic A: Validation of a primer optimisation matrix toimprove the performance of reverse transcription-quantitative real-timePCR assays. BMC Res Notes 2009, 2:112.
25. Van der Stoep N, van Paridon CDM, Janssens T, Krenkova P, Stambergova A,Macek M, Matthijs G, Bakker E: Diagnostic Guidelines for High-ResolutionMelting Curve (HRM) Analysis: An Interlaboratory Validation of BRCA1Mutation Scanning Using the 96-Well LightScannerTM. Hum Mutat 2009,30:899-909.
26. De Leeneer K, Coene I, Poppe B, De Paepe A, Claes K: Rapid and SensitiveDetection of BRCA1/2 Mutations in a Diagnostic Setting: Comparison ofTwo High-Resolution Melting Platforms. Clin Chem 2008, 54:982-989.
27. de Juan I, Esteban E, Palanca S, Barragán E, Bolufer P: High-resolutionmelting analysis for rapid screening of BRCA1 and BRCA2 Spanishmutations. Breast Cancer Res Treat 2009, 115:405-414.
28. Kennerson ML, Warburton T, Nelis E, Brewer M, Polly P, De Jonghe P,Timmerman V, Nicholson GA: Mutation scanning the GJB1 gene withhigh-resolution melting analysis: implications for mutation scanning ofgenes for Charcot-Marie-Tooth disease. Clin Chem 2007, 53:349-352.
29. Martino A, Mancuso T, Rossi AM: Application of high-resolution melting tolarge-scale, high-throughput SNP genotyping: a comparison with theTaqMan method. J Biomol Screen 2010, 15:623-629.
30. Montgomery JL, Sanford LN, Wittwer CT: High-resolution DNA meltinganalysis in clinical research and diagnostics. Expert Rev Mol Diagn 2010,10:219-240.
31. Wagner T, Stoppa-Lyonnet D, Fleischmann E, Muhr D, Pagès S, Sandberg T,Caux V, Moeslinger R, Langbauer G, Borg A, Oefner P: Denaturing high-performance liquid chromatography detects reliably BRCA1 and BRCA2mutations. Genomics 1999, 62:369-376.
Hondow et al. BMC Cancer 2011, 11:265http://www.biomedcentral.com/1471-2407/11/265
32. Dufresne SD, Belloni DR, Wells WA, Tsongalis GJ: BRCA1 and BRCA2mutation screening using SmartCycler II high-resolution melt curveanalysis. Arch Pathol Lab Med 2006, 130:185-187.
33. Takano EA, Mitchell G, Fox SB, Dobrovic A: Rapid detection of carriers withBRCA1 and BRCA2 mutations using high resolution melting analysis.BMC Cancer 2008, 8:59.
34. Vorkas PA, Christopoulos K, Kroupis C, Lianidou ES: Mutation Scanning ofexon 20 of the BRCA1 gene by high-resolution melting curve analysis.Clin Biochem 2009, 43:178-185.
35. Kwong A, Ng EKO, Law FBF, Wong LP, To MY, Cheung MT, Wong HN,Chan VW, Kurian A, West DW, Ford JM, Ma ESK: High-resolution meltinganalysis for rapid screening of BRCA2 founder mutations in SouthernChinese breast cancer patients. Breast Cancer Res Treat 2009, 122:605-607.
36. Coulet F, Pires F, Rouleau E, Lefol C, Martin S, Colas C, Cohen-Haguenauer O, Giurga I, Fajac A, Nogue’s C, Demange L, Hardouin A,Lidereau R, Soubrier F: A One-Step Prescreening for Point Mutations andLarge Rearrangement in BRCA1 and BRCA2 Genes Using QuantitativePolymerase Chain Reaction and High-Resolution Melting Curve Analysis.Genet Test Mol Biomarkers 2010, 14:677-690.
37. Zhou L, Wang L, Palais R, Pryor R, Wittwer CT: High-resolution DNAmelting analysis for simultaneous mutation scanning and genotyping insolution. Clin Chem 2005, 51:1770-1777.
38. De Leeneer K, Coene I, Poppe B, De Paepe A, Claes K: Technical AdvanceGenotyping of Frequent BRCA1/2 SNPs with Unlabeled Probes. J MolDiagn 2009, 11:415-419.
39. Krypuy M, Newnham GM, Thomas DM, Conron M, Dobrovic A: Highresolution melting analysis for the rapid and sensitive detection ofmutations in clinical samples: KRAS codon 12 and 13 mutations in non-small cell lung cancer. BMC Cancer 2006, 6:295.
40. Krypuy M, Ahmed AA, Etemadmoghadam D, Hyland SJ, DeFazio A, Fox SB,Brenton JD, Bowtell DD, Dobrovic A: High resolution melting for mutationscanning of TP53 exons 5-8. BMC Cancer 2007, 7:168.
Pre-publication historyThe pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2407/11/265/prepub
doi:10.1186/1471-2407-11-265Cite this article as: Hondow et al.: A high-throughput protocol formutation scanning of the BRCA1 and BRCA2 genes. BMC Cancer 201111:265.
Submit your next manuscript to BioMed Centraland take full advantage of:
• Convenient online submission
• Thorough peer review
• No space constraints or color figure charges
• Immediate publication on acceptance
• Inclusion in PubMed, CAS, Scopus and Google Scholar
• Research which is freely available for redistribution
Submit your manuscript at www.biomedcentral.com/submit
Hondow et al. BMC Cancer 2011, 11:265http://www.biomedcentral.com/1471-2407/11/265
